Beruflich Dokumente
Kultur Dokumente
ISSN: 2321-8169
Volume: 3 Issue: 8
5322 - 5327
__________________________________________________________________________________________________________________________________
Classification and Grading of Wheat Granules using SVM and Naive Bayes
Classifier
Ms. Rupali S. Zambre
Research Scholar,
G.H.R.I.E.M., Jalgaon
zambrerupali@gmail.com
Assistant Professor,
G.H.R.I.E.M., Jalgaon
sonalpatil3@gmail.com
Assistant Professor,
G.H.R.I.E.M., Jalgaon
ganesh.dhanokar@raisoni.net
Abstract- India is the second leading producer of wheat in the world. Specifying the quality of wheat manually is very time consuming and
requires an expert judgment. With the help of image processing techniques, a system can be made to avoid the human inspection. Classification
of wheat grains is carried out according to their grades to determine the quality. Images are acquired for wheat grains using digital camera.
Conversions to gray scale, Smoothing, Thresholding, Canny edge detection are the checks that are performed on the acquired image using image
processing technique. Classification and Grading of wheat grain is carried out by extracting morphological, color and texture features. These
features are given to SVM and Naive Bayes Classifier for classification. To evaluate the classification accuracy, from the total of 1300 data sets
50% were used for training and the remaining 50% was used for testing. The classification system was supervised corresponding to the
predefined classes of grades. Results showed that overall accuracy of SVM and Naive Bayes classifier is 94.45%, 92.60% respectively. So, the
classification performance of SVM is better than Naive Bayes Classifier.
Keyword: Smoothing; Thresholding; Canny Edge detection; Support Vector Machine; Naive Bayes Classifier.
__________________________
___________________*****________________________________________
I. INTRODUCTION
Wheat is the very important and widely grown food crop in
the world. India is the second largest producer of wheat in
the world because of its valuable nutrients. It is the main
food resource of protein, energy and dietary fiber in human
nutrition and an important industrial raw material as well.
Grain grading system gives assurance that a
particular lot of grain meets the required set standards of
customers. The quality of wheat has different effect on the
yield of wheat; therefore it is important to have the proper
inspection of wheat quality. In many countries, grading of
grain depends on some properties; test weight, moisture
contents, broken foreign material or the percentage
fragments example broken corn foreign materials, damaged
kernels. These grains consist of several impurities like
stones, weed seeds, damaged seeds, more broken granules
etc [1].
In the present system, grain type and quality are
fastly assessed by visual inspection. Determining the quality
of wheat manually is time consuming and requires an expert
judgment. The decision-making capabilities of a grain
supervisor can be critically affected by his/her physical
condition, working conditions such as climate, improper
lighting etc and loyalty for traders. Also, the knowledge and
experience of inspectors are required to accurately perform
this evaluation process. Thus, these tasks require automation
and develop imaging systems that can be supportive to grade
wheat grains. Different algorithms can be used to classify
wheat according to its quality. In this paper, the proposed
work contains the Support Vector Machine and Naive Bayes
Classifier for classifying the wheat grains according to their
grades.
_______________________________________________________________________________________________
___
B. Image Preprocessing
The raw data is subjected to several preliminary processing
steps to make it functional in the descriptive stages of
classification and grading. In order to get wheat grain
features accurately, wheat grain images are preprocessed
through different preprocessing methods.
First the image will be converted into gray scale
image. Generally, the image will pre-processed using a
smoothing filter that include operations which can enhance
images and remove noise from an image [4]. The noise
present into the input image is removed by using median
filter.
Conversion of a gray-scale image into a binary
image is called as thresholding or binarization. The output
binary image has values of 1 (white) for all pixels in the
input image and 0 (black) for all other pixels. Thresholding
separate the wheat kernel from the background by
converting the image into binary image. Thresholding is also
used for segmentation which subdivides an image into
different parts or objects i.e. the objects of interest are
isolated from their background. Otsu method is used for
binarization. Due to some defects on the surface of wheat
grains or over drying of the wheat grain, there exist holes.
These holes were filled with neighboring pixels.
C. Edge Detection Technique
A. Image Acquisition
The Initial step in using a machine vision system is to obtain
a digital image. Proper light plays important role in
obtaining a good image. This can cause alteration of object
features in the image. Determining ideal illumination source
is not easy and it depends on the nature [9]. Here, the
samples of Wheat grains are collected from local farmers.
Wheat grain images are captured using a digital camera. A
fixed distance between the camera and the grain samples is
maintained. A uniform background which is black in color.
The grains are spread on a black sheet randomly. Although
grains are placed randomly, make sure that they are not in
contact with other. The images were captured and stored in
hard disk in JPEG (Joint Photographers Expert Groups)
format. The basic building block for image capturing is
shown in Fig. 2.
_______________________________________________________________________________________________
___
The denominator
in the above
_______________________________________________________________________________________________
___
No of Images
Classified
Correctly
No of Images
Classified
Incorrect
Grade 1
130
121
Grade 2
130
119
11
Grade 3
130
118
12
Grade 4
130
122
Grade 5
130
123
By SVM
Class
(Grade)
No of
Images
Tested
No of Images
Classified
Correctly
No of Images
Classified
Incorrect
Grade 1
130
125
Grade 2
130
119
11
Grade 3
130
120
10
Grade 4
130
126
Grade 5
130
124
Class
(Grade)
Classifie
r
Grad
e1
Grade
2
Grad
e3
Grad
e4
Grad
e5
Overall
Accu
racy
96.15
%
91.53
%
92.30
%
96.92
%
95.38
%
94.45
%
92.30
%
91.53
%
90.76
%
93.84
%
94.61
%
92.60
%
SVM
Naive
Bayes
Classifie
r
5325
IJRITCC | August 2015, Available @ http://www.ijritcc.org
_______________________________________________________________________________________________
___
V. DISCUSSION
The experiments are conducted by using same feature sets as
input to both SVM and Naive Bayes Classifier. Then, the
experiment results are compared the performance of the
SVM and Naive Bayes classifier. No standard database was
available so, created the dataset containing total 1300
images of wheat grain. The 50% dataset i.e. 650 images
used for training and 50% dataset i.e. 650 images used for
testing. The accuracy of classification varies differently for
different classifiers The morphological, color and textural
features are extracted and same feature set is given to
Support Vector Machine and Naive Bayes Classifier for
classification. The performance result i.e. accuracy of each
grade for both classifiers i.e. for Support Vector Machine
and Naive Bayes Classifier is shown in Table 3 and in Fig.
3. The SVM perform better than Naive Bayes Classifier for
Grade 1, Grade 3, Grade 4, and Grade 5 while the
classification result for Grade 2 is same for both SVM and
Naive Bayes Classifier. The overall accuracy of SVM and
Naive Bayes is 94.45% and 92.60% respectively. The
Overall accuracy for classifying wheat grains in to grade by
_______________________________________________________________________________________________
___
[10]
[11]
[12]
[13]
5327
IJRITCC | August 2015, Available @ http://www.ijritcc.org
_______________________________________________________________________________________________
___